Jump to content

The Death of Mesh


Lilith Heart
 Share

You are about to reply to a thread that has been inactive for 4605 days.

Please take a moment to consider if this thread is worth bumping.

Recommended Posts

Why stop the downscaling there? I could use some of the content I've developed for mobile use outside Second Life.

You can easily fit 50+ buildings on one 512 texture and use less than 1000 triangle for the batch.

50+ animated objects using another single 512 with another maybe 2000 triangle & a one liner script to contol them all.

Simple avatars that are only a handful of pixels in any dimension (with careful camera controls so nobody zooms in to the point everything looks like ****)

Like this, you could fit a high-density city usable by micro-avatars on a single sim. Ultracheap :)

Maybe this is the kind of content application Lindens hope to see?

Link to comment
Share on other sites

How disappointing. I was away from SL for... what, nearly two years now, and mesh tempted me back. I've been messing around in Blender the past few days. And now I find that PE makes it effectively unusable? Sigh. Look, I know meshes have a streaming cost higher than sculpties and especially prims. Making them cost a bit more than sculpties or regular prims? Okay, fine. I won't say I like it, considering the savings you can get in polygons and texture sizes, but it's not unreasonable. Twenty times more is not "a bit." 125 times more is definitely not "a bit." Why even bother developing mesh if it's going to be so costly, PE wise?

Link to comment
Share on other sites

PE just starts to get into more reasonable range, at least for medium sized objects ( smaller than 10 meters). In that size range i am able to get very close to the costs (PE wise) compared with builds made out of Sculpted prims (as long as i compare a build made of 3 Sculpties or more with an equivalent mesh build).

Link to comment
Share on other sites

hey Gaia, I really don't think anything 10m or smaller could be described as a medium sized object. 10m is a small object. Medium sized objects start at about 40m and large would be 60m +

In my tests for buiding furniture and other small objects , less than 10m, sculpties are still winning hands down for prim costs.

Sculpties should not be winning over mesh in any size range, it makes mesh pretty pointless, except for AV products

Link to comment
Share on other sites

i disagree in two ways:

 

  1. I think that the wast majority of objects is in the range from 1 cm to 10 meters. So i consider bigger objects as ... big. despite that you now CAN have objects of sizes up to 64 meters (bbox wise).
  2. In my tests with objects up to about 2 meters i can get my objects into the same range of PE  regardless whether they are made out of meshes or sculpties or prims. this is true as long as the objects contain more than 2 prims. And the more prims they actually contain the better mesh performs. So (as vivienne daguerre has shown on another thread) even houses (which are certainly big objects...)can become cheaper when meshes are used.

For me using meshes (in the given range of size) gives me following benefits

  1. I have much more control over modelling. So i can mold my prims to my needs and can potentially do high optimizations.
  2. I have much more control over texturing.
  3. I have 8 times more texturable faces, so i can do much more detailed texturing.
  4. Because i can split up my textures into up to 8 individual images per prim, i can make versatile use of tiled textures and such i can reduce the amount of needed texture data enormously.
  5. I have immediate control over physics. No longer do i need to add invisble prims and set my object to phantom. This works well as long as i provide an easy basic physics shape.
  6. I have full control over LOD. I can do high optimizations especially for lowest LOD, which makes a big difference for objects in the examined range (1cm - 10 meters)

The backdraws with meshes (true for any size) for me are:

 

  1. It takes me a lot of more work to define the LOD meshes
  2. I have significant more work to keep the unwraps of the LOD meshes intact. I do not use decimate, but i make LOD optimizations by hand. This is the only way by which i can get acceptable LOD behaviour (well i am very picky here)
  3. Changes on ready made Objects are painfull because of the changes i have to propagate into the unwraps.
  4. Texturing work raises up significantly.

Taking upload fee aside for a moment, i am starting to get confident that meshes can compete well with sculpties as long as high quality visual appearance is of concern (and as long as we talk about the given range of size).

So wherever you want something to look realy nice, meshes look like they are very competitive. For Low prim/low budget work sculpties will still win.

Link to comment
Share on other sites

"I have significant more work to keep the unwraps of the LOD meshes intact. I do not use decimate, but i make LOD optimizations by hand. This is the only way by which i can get acceptable LOD behaviour "

Oh yes. I agree with this so much. If only we could have a script so that the changes made to the mesh with Alt+M were automatically propagated to the UV map!



Link to comment
Share on other sites

  • 4 weeks later...

I had such high hopes for mesh.  Meshes do free you from the incredibly cumbersome aspect ratio restraints sculpts require.  But when I see a 25m sculpted tree show as 3 prims only to become 50 even when uploaded as an enhanced efficiency mesh using fewer faces I get incredibly discouraged.   What's the point of building for mesh with false equivalences like this?

Link to comment
Share on other sites

Thats one side of the coin for sure, i think scuplts will be useful in alot of situations like trees for example, but i think meshes are as useful in lots of other areas. In most cases where scuplts are not useful (like large builds) meshes are competing very well in terms of prim counts. I cant really say im dissappointed with it really, my current project is standing at 50PE and i have no doubt at all that if i did the same with prims it would be 4 times more at least.

Link to comment
Share on other sites

Spent a good part of the day reading up on the PE issue and it seems clear that LL doesn't want big builds using mesh.  I guess that makes sense.   It's not hard to imagine what might happen if people could upload huge mesh items (without the constraints PE enforces).  

That said, it doesn't make much sense to see a 20m mesh build using 2/3 the faces of like sculpts and have the mesh produce an 85 PE tree when the sculpt tree uses 5 prims at any scale.  Even the least efficient 1 sculpt per tree limb, branch, trunk approach produces a tree with less than 25 prims and at any scale.   Seems a bit crazy and way outdated to make meshes so much less efficient, ie high PE ratios.  Personally, seems like a fixed, non-scaleable mesh/size alternative makes sense.

The other issue I've seen argued regarding PE is in selling larger mesh items.  How do you sell a scaleable 10m, 20 PE tree to someone who then scales the tree to 25m only to find they've used 80 prims on their prim limited land without getting questions, complaints, etc?  That issue alone, absent grid wide user education, seems to relegate meshes to niche for large item builders.

It is what it is.  I'm back to sculpteds having wasted a week working out the (in)efficiencies of meshes.

Link to comment
Share on other sites

" doesn't make much sense to see a 20m mesh build using 2/3 the faces of like sculpts and have the mesh produce an 85 PE tree when the sculpt tree uses 5 prims at any scale.  Even the least efficient 1 sculpt per tree limb, branch, trunk approach produces a tree with less than 25 prims and at any scale.  "

Sculpts are a very useful legacy building option.  I'd say to be able just to  make a sculpted tree  is a gift and  its amazing that we can have a mesh  tree using only 5 prims.   One of the reasons 3d game artists use matt painted backrounds is a trade off issue.  A tree, unless its a main feature, like a tree house, is a backround item, best made from images projected on a plane using alphas or included in a backdrop painting. 

A tree is highly complex and completely made up of complex curves.  If you are able to view it from every possible angle, it has to be a tremendous load on the system.  Every curve in you mesh multiplies the complexity of that area of  mesh at least 3x (I'm bad at math, maybe it's more - but you need to add about 3 edgeloops to begin to make it smooth)  Sharp corners simple take less polys.

Link to comment
Share on other sites

Not sure what you're saying Nacy.  Most of the MMO games I'm familiar with use object trees.   WoW is mostly old tech whereas Rift is quite new and its trees and foliage reflect that enhanced state.  So, I'm not sure where the matte or flat panel references you mention fit in.  Even SL has dimensional trees (licensed from Speedtree from 2006 or before?).

I'm just saying that that from a render cost standpoint a 25 prim, fully scalable, sculpted tree doesn't seem to make sense relative to a 5 prim mesh version, particularly where each mesh object has fewer faces than any of the individual 25 sculpted prims.  Why should the mesh tree have an 85 PE then?

Maybe I'm missing something.  What's the difference in rendering cost between identically shaped and scaled objects when one uses a sculpted texture and another is a mesh?   If there is one and it's significant, then that probably explains the PE difference.  If not, or negligible, then I kind of assume the PE is about preventing willy nilly 2 million polygon mesh uploads.

Like I was saying.   It is what it is until LL decides to change it.  It's a big disadvantage to be stuck with sculpts however.

Link to comment
Share on other sites

You are about to reply to a thread that has been inactive for 4605 days.

Please take a moment to consider if this thread is worth bumping.

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
 Share

×
×
  • Create New...